Download (2002), "An on-line look at sentence processing in the second

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Lexical semantics wikipedia , lookup

Spanish grammar wikipedia , lookup

Pleonasm wikipedia , lookup

Transformational grammar wikipedia , lookup

Pipil grammar wikipedia , lookup

Musical syntax wikipedia , lookup

Junction Grammar wikipedia , lookup

Parsing wikipedia , lookup

Transcript
Frenck-Mestre
On-line sentence processing in the second language
In J. Altarriba and R. Herridia (Eds.) Syntactic processing in the second language
Frenck-Mestre, C. (2002). A on-line look at sentence processing in a second language. In R.
Herrida and J. Altarriba (Eds.) Bilingual Sentence Processing. North Holland
An on-line look at sentence processing in the second language
C. Frenck-Mestre
What are the factors that affect immediate syntactic processing in the second
language of bilinguals? How might processing evolve with increasing skill in the
second language? Herein, we will review these questions in the light of data from
several on-line investigations of syntactic processing in the native and second
language of bilinguals with varying levels of second language proficiency.
Performance is examined, moreover, for bilinguals of different language
backgrounds. The data from these experiments highlight the role of L2 exposure, in
line with models based upon linguistic experience.
The present chapter is concerned with the acquisition of syntax in the second
language by adult learners, a topic that has received considerable attention and
fuelled many a debate. We will review the handful of recent on-line psycholinguistic
studies of second language syntactic processing in adult, late bilinguals, in order to
pinpoint the factors that influence immediate syntactic analysis. None would
dispute the claim that the final comprehension of a sentence is affected by a myriad
of factors (at the lexical, syntactic, discourse and pragmatic levels). Our aim here,
however, is to provide a clear picture of the factors that play an immediate role in
second language syntactic analysis. We will examine the results obtained in studies
of adult bilinguals at different levels of second-language proficiency, as well as
different language pairs, to establish the role of exposure to the second language in
acquiring L2 syntax and the influence of "forward transfer." As we will show, there
is substantial evidence, first, that "exposure-based" models such as that proposed by
Mitchell and colleagues (Cuetos, Mitchell & Corely, 1996; Mitchell, 1994) may
provide a viable framework for understanding how adults aquired the syntax of a
second language, and second, that many common processes underlie native and
second-language syntactic processing.
Why is second-language reading slow?
Quite often, on-line measurements obtained during the reading of single
sentences reveal increased processing time for non-native readers (cf. Fernandez,
2000; Frenck-Mestre, 1997; Frenck-Mestre & Pynte, 1997; Hoover & Dwivedi,
1998; Seglaowitz & Herbert, 1990). In the absence of specific lexical and/or
syntactic "deficits," i.e. in the case that the second language reader has access to
both the lexical and syntactic information present in the sentence, the underlying
cause for this relative slowness is not immediately apparent. It has been suggested
that a lesser automatization of lower-level processes may be responsible. For
example, a lesser ability to encode orthographic redundancies may slow lexical
access in the second language (Favreau & Segalowitz, 1983, but see Frenck-Mestre,
1993). This would not only incur a general slowing but also a more specific
difficulty, related to phonological activation. That is, given what we know about the
time course of orthographic and phonological activation (see, for example, Ferrand
& Grainger, 1993; Frost, Katz & Bentin, 1987; Paap, McDonald, Schvaneveldt &
Noel, 1987), slower orthographic processing would allow for greater phonological
activation and hence possible interference when accessing homophones (such as, in
On-line sentence processing in the second language
English, "fare" versus "fair" "great" versus "grate" etc. ). Indeed, just such a result
was reported by Segalowitz and Hébert (1990); bilinguals who read slower in their
second than native language were more prone to interference from homophones
when processing semantically anomalous sentences (e.g. "The weather was fare.")
than were bilinguals were read equally fast in their two languages. While second
language reading may indeed suffer from a lesser degree of automaticity as concerns
such encoding processes, experiments performed in our laboratory provide evidence
that other factors may also be at play.
If one records the eye movements of (non-native or native) readers, it is possible
to distinguish the "first pass" through the sentence from eventual re-readings,
whether of the entire sentence or parts thereof (see Rayner & Pollatsek, 1987;
Rayner, Sereno, Morris, Schmauder & Clifton, 1989, for discussions of eyemovement recording and its use in psycholinguistic studies). This tool has enabled
our research group to show that at a gross level, increased reading times for
proficient non-native readers is often linked to a tendency to re-read sentences. This
has been observed informally, in numerous experiments involving diverse syntactic
structures, and reported in various talks and published studies (Frenck-Mestre,
1998a, 1998b; Frenck-Mestre & Pynte, 1995; 1997). It is worth restating that this
general tendency to reread sentences is observed for proficient bilinguals, i.e. in
spite of a high level of lexical and syntactic knowledge in the second language. A
different picture arises when we consider the oculomotor behaviour of these
bilinguals during the "first pass" through second language sentences (i.e. in the
course of the first left-to-right pass through it). For first pass measures, advanced
non-native readers and native readers do not differ notably in their performance at
least insofar as we have observed (Frenck-Mestre, 1998a, 1998b; Frenck-Mestre &
Pynte, 1995; 1997). That is, we have not found the second language performance of
these speakers to differ notably from native performance as concerns the average
length of a saccade, the mean duration of fixations, or the probability of making a
regressive saccade during the first reading of the sentence. It is perhaps important to
underline the differences, here, between the information provided by eye-movement
recording as opposed to self-paced reading. While the latter technique allows one to
track on-line processing, it does not allow one to distinguish between "first" versus
"second" readings of parts of the sentence. It may be that "rereading" of the
segments presented in self-paced reading studies, rather than initial processing as
concerns either lexical access or syntactic analysis, are also the reason behind slower
reaction times reported in these studies.
How then should one interpret "slow" non-native reading? Is it perhaps linked
after all to differences and/or "deficits" in syntactic processing as compared to native
readers' processing? Indeed, one could hypothesize that, despite their second
language proficiency, even skilled bilinguals do not perform a full syntactic analysis
during the first pass through the sentence, thus rendering necessary a second pass
through it. As we report in greater detail later in this chapter, however, this
hypothesis is untenable in the face of other on-line results. Indeed, proficient
second-language readers demonstrate an immediate sensitivity to the same factors
that influence native readers' initial progression through the sentence. Both
structural ambiguity and lexical constraints influence the first pass reading times of
proficient non-native readers, just as they do those of native readers (Frenck-Mestre,
1998a, 1998b; Frenck-Mestre & Pynte, 1997). It can be noted that the same
immediate effect of syntactic ambiguity on second language processing has been
reported using self-paced reading (Hoover & Dwivedi, 1998; Juffs & Harrington,
1996). Hence, slower second language reading times, at least in skilled bilinguals,
do not appear to be linked to differences in syntactic processing as compared to
native readers (see also Segalowitz, 1986; Segalowitz & Hébert, 1990), as will be
discussed shortly. Our own hypothesis, indeed which remains to be tested through
systematic study, is that the tendency to re-read may be a "residual" of early
Frenck-Mestre
difficulties experienced in second language reading. That is, at an early stage, rereading in the second language may well be necessary due to lesser abilities (as
concerns orthographic encoding, lexical access and/or syntactic analysis). Later on,
the more skilled bilingual may retain this "developed habit" of re-reading, despite it
no longer being necessary. Although this remains to be demonstrated, it is in line
with the results reported below showing similar syntactic processing for native
speakers and skilled bilinguals in their second language, despite longer overall
reading times in the latter.
Slower but not necessarily different syntactic processing in the second language
Consider, first, three recent studies that examined the on-line processing of
syntactically ambiguous sentences, in both native and second language readers. In
two of these studies, the performance of proficient English-French bilinguals
revealed immediate sensitivity to syntactic ambiguity quite independently of their
overall reading speed. In the third, however, the pattern of results from Chinese ESL
learners, who were both less proficient than the bilinguals who participated in the
other two studies and quite slow as compared to native speakers, is more
questionable as concerns their sensitivity to second language syntactic ambiguity.
Hoover and Dwivedi (1998) looked at the on-line processing of a syntactic
ambiguity that is present in romance languages, in the current case French, but
absent in English. The ambiguity is related to the role of the clitic pronoun
(highlighted in the example) in French causative sentences, as illustrated below:
1a. Sarah le fera signer en présence de son avocat.
(Sara it will have signed in the presence of her lawyer)
1b. Sarah le fera demain en présence de son avocat.
(Sara it will do tomorrow in the presence of her lawyer).
In both 1a and 1b, a full direct object NP has been pronominalised and replaced
by the clitic pronoun "le." However, in the first example, 1a, the clitic is ambiguous.
This is because the verb "faire" can either be used as a thematic verb, as is the case
in 1b, or in a causative construction as illustrated in 1a. Therefore, upon reading the
clitic in 1a, the reader could initially treat it as an argument of the proceeding verb
"faire" only to find this an erroneous assignment upon the reading of the second verb
(in the example, "signer"). In 1b, no ambiguity is present, as the clitic pronoun can
only be assigned as an argument of the verb "faire." This ambiguity is known to
cause difficulty for native French speakers, as was demonstrated experimentally in a
self-paced reading study by Dwivedi & Hoover (1996, reported in Hoover &
Dwivedi, 1998). In a subsequent self-paced reading study, Hoover and Dwivedi
(1998) looked at the processing of this ambiguity for two groups of highly fluent L2
French readers (L1 English). While both groups rated quite high as concerns level
of proficiency in French, one group was classified as "fast" readers on a reading test
and the other comparatively "slow." The main objective of the study was to
determine whether sensitivity to the ambiguity present in French clitic/causative
constructions would be correlated with reading speed in the second language. That
is, whether "slow" L2 readers would be less sensitive to structural ambiguity than
"fast" readers. Quite clearly, the answer to this question was negative. First, it can
be noted that both groups of L2 readers were significantly slower than native
readers. Despite this, no interactions between reading group (native, fast L2 and
slow L2) and ambiguity resolution were observed at the point in the sentence where
disambiguation occurred. Both slow and fast readers showed immediate sensitivity
to structural ambiguity in the L2, in like fashion to native readers. That is, all groups
took significantly longer to read the second verb in causative/clitic sentences such as
On-line sentence processing in the second language
1a than in unambiguous constructions. Thus, in this study reading speed per se was
not indicative of differences in syntactic processing, either between L2 readers
themselves or as compared to native speakers.
A similar pattern of results was reported by Frenck-Mestre & Pynte (1997,
experiment 2), in an eye-movement study involving proficient English-French and
French-English bilinguals. These bilingual subjects displayed slower overall
reading times in their second as compared to native language, and as compared to
native speakers, however this factor was not correlated with differences in syntactic
processing as compared to native processing. The results of this study showed that
not only are L2 readers sensitive to syntactic ambiguities in their second language,
but moreover they can use lexical subcategorisation information to resolve certain
structural ambiguities. Consider the example provided in 2, below
2a. Whenever the dog obeyed the little girl she showed her approval.
2b. Whenever the dog obeyed the little girl showed her approval.
2c. Whenever the dog barked the little girl showed her approval
The first of the above examples, 2a, poses less difficulty than the second, 2b, for
native speakers (Rayner, Carlson & Frazier, 1983). This is due, according to one
theoretical framework, to the parser systematically adopting an initial direct object
analysis of the second NP, only to find this inappropriate in the case of 2b upon the
reading of the second verb (Frazier, 1987). Reanalysis is thus necessary in 2b, which
is costly. In addition, in this framework 2b and 2c should both cause the parser to
initially adopt a direct object NP analysis, although this error would be quickly
recognized in 2c due to the lexical properties of the subordinate verb coming into
play. As such, 2c is also easier to parse than 2b (Frenck-Mestre & Pynte, 1995;
Mitchell, 1989; Mitchell & Holmes, 1985; Trueswell, Tanenhaus & Kello, 1993).
The same result was obtained for proficient French-English and English-French
bilinguals reading in their second language (Frenck-Mestre & Pynte, 1997,
experiment 2). Sentences such as 2c were processed faster, causing fewer rereadings and fewer regressive saccades, than sentences such as 2b. Otherwise
stated, these L2 readers were "led up the garden path" in the case of sentences like
2b just as native readers are, whereas in the case of 2c the subcategorisation
information present in the subordinate verb helped these L2 readers to quickly
resolve this syntactic ambiguity, just as it has been shown in native readers.
Another recent on-line study of L2 syntactic processing examined the same
ambiguity as illustrated in 2b and 2c, but in less-proficient bilinguals. Juffs and
Harrington (1996) looked at Chinese ESL students' processing of this structural
ambiguity by means of a word-by-word self-paced reading experiment. The authors
themselves concluded that their ESL readers were sensitive to the syntactic
ambiguity present in these structures, and able to resolve the ambiguity faster for
sentences in which the subordinate verb was intransitive (2c) than optionally
transitive (2b). Indeed, these L2 readers did experience apparent difficulty with the
matrix verb in sentences such as 2b as revealed by an abrupt increase in reading
times at the matrix verb for these sentences. Moreover, this abrupt increase was not
observed at the matrix verb in sentences such as 2c, but one constituent earlier, i.e. at
the NP following the intransitive subordinate verb, thus indicating that the
subcategorisation information provided by the subordinate verb was used to
block/re-analyse a direct-object analysis of the NP. However, some caution may be
warranted prior to concluding that these ESL readers had lesser difficulty with the
structural ambiguity in the intransitive than optionally transitive cases. In fact, this
direct comparison was not made in the study. The inspection of means does not lead
one to believe, moreover, that at the matrix verb itself any significant interaction
would obtain in the ESL group. As such, the conclusions drawn by Juffs and
Harrington that ESL readers experienced the same processing difficulty as native
Frenck-Mestre
speakers with "garden path" sentences may need to be nuanced.
Yet another eye-movement experiment, reported in Frenck-Mestre and Pynte
(1997, exp. 1) revealed quite impressive use of subcategorisation information in the
resolving of structural ambiguities by second-language readers. The ambiguity
studied therein involved the attachment of a prepositional phrase. Consider, first,
monolingual studies of this ambiguity, illustrated in 3.
3a. The spy saw the cop with the binoculars but the cop didn’t see him.
3b. The spy saw the cop with the revolver but the cop didn’t see him.
In an early eye-movement study with monolinguals, Rayner, Carlson & Frazier
(1983) demonstrated that readers systematically prefer 3a to 3b. That is, readers had
greater difficulty (reflected by increased reading times) when the sentence biased
“low” attachment of the prepositional phrase, to the preceding noun phrase (“with
the revolver” is associated with “the cop”), than when it biased toward “high”
attachment to the preceding verb phrase (“with the binoculars” thus modifies the
verb “to see”). In one theoretical framework, this preference to attach “high” is
attributed to an autonomous parser which obeys heuristic principles and prefers the
least costly structure in terms of syntactic nodes (Frazier, 1987). The parser is
assumed to operate independently of other levels of analysis – lexical, pragmatic,
and/or referential – as attested by the fact that even when semantic/pragmatic factors
favor “low” attachment for this structure, as in 3b (cops are known to carry
revolvers) the initial analysis of the sentence, as revealed by first pass reading times,
still reflected a “high attachment preference.”
The conclusions from this seminal study were brought into question, however,
by subsequent monolingual studies in which the influence of thematic constraints
and referential context were clearly demonstrated (Altmann & Steedman, 1988;
Taraban & McClelland, 1988). Parsing preferences were not found in these on-line
(self paced reading) studies to be systematic, but to vary as a function of these
“extra-syntactic” factors. Low attachment was shown to be preferred when either
thematic constraints or the referential context disambiguated toward this structure.
In addition, the lexical constraints of the initial VP have been shown to immediately
influence the processing of this ambiguity in the native language of readers (FrenckMestre & Pynte, 1997; see also Boland & Boehm-Jernigan, 1998 for similar work
on PP attachment, but Ferreira & Henderson, 1990, for counter-arguments). The
same is true of proficient English-French bilinguals reading in their second language
(Frenck-Mestre & Pynte, 1997). Consider the examples provided in 4.
4a. Il rate le train de nuit et décide alors de chercher un hôtel.
(He missed the train of night and decided to look for a hotel)
4b. Il rate le train de peu et décide alors de chercher un hôtel.
(He missed the train by little and decided to look for a hotel)
4c. Il avertit la police du quartier puis se félicite de son action.
(He warned the police of the district and congratulated himself for it.)
4d. Il avertit la police du crime puis se félicite de son action.
(He warned the police of the crime and congratulated himself for it.)
In 4, the bias towards « high » or « low » attachment of the ambiguous PP
(highlighted in the examples) is provided by the lexical constraints of the verb
preceding it. In 4a and 4b, the matrix verb is monotransitive, i.e. it generally selects
for a single direct object complement. For this class of verb, then, the immediately
following NP can satisfy the thematic grid. In contrast, the matrix verb in 4c and 4d
is ditransitive, that is it generally selects for two complements, only one of which is
provided by the post verbal NP. For this class of verb a second complement is
On-line sentence processing in the second language
expected, which can be satisfied by the PP. This type of subcategorisation
information influences the immediate decisions made by the parser (Frenck-Mestre
& Pynte, 1997). For monotransitive verbs, the immediate preference of native
French readers--as shown both off-line (via sentence completion data) and on-line
(via the recording of eye movements during reading)--is to attach low, to the
preceding NP. That is, at the disambiguating element (the noun of the PP) readers
experience less difficulty with 4a than with 4b. The exact opposite is observed for
sentences containing ditransitive verbs. In this instance, first pass reading times are
longer at the disambiguating element for low than high attachment of the PP, i.e. for
4c than 4d.
Fluent bilingual readers show the same sensitivity to lexical constraints when
resolving the syntactic ambiguity illustrated in 4. English native speakers reading in
French prefer low attachment of the PP for monotransitive verbs (4a as compared to
4b), but high attachment following ditransitive verbs (4d as compared to 4c).
Moreover, this effect is observed on first pass reading times, as it is in native
speakers. Indeed, the pattern of processing did not differ as a function of “type of
reader,” i.e. whether first or second language.
What can be concluded from the above set of studies? Quite clearly, fluent
bilinguals are not only sensitive to structural ambiguities when reading in their L2,
but moreover are sensitive to fine levels of analysis. Although we did highlight some
important differences across these three studies, it is more likely, in our opinion, that
the difference in sensitivity to syntactic ambiguity in the second language across
these three studies is related to L2 proficiency rather than simply reading speed.
Slower second language reading is not necessarily indicative of different processing
as compared to native speakers, as shown by Hoover and Dwivedi (1998) as well as
by Frenck-Mestre and Pynte (1997). However, level of proficiency indeed seems to
be an important factor, as can be seen in the Juffs and Harrington (1996) study.
Less-proficient bilinguals may, first, be unable to immediately process syntactic
ambiguity in their second language. Second, as we will outline in the next section,
when a structural ambiguity can be resolved differently across languages, nonproficient bilinguals may be susceptible to effects of forward transfer. Hence, they
may process the ambiguity differently from native speakers.
The role of experience (years of exposure) and effects of transfer in L2 parsing
In the previous section we reviewed three on-line studies where the native
language of the bilingual readers was generally not considered to inhibit second
language syntactic processing. In those studies, either the structure was absent from
the native language of subjects (cf. Hoover & Dwivedi, 1998) or it was both
identical across the bilinguals' two languages and resolved in similar manner in the
two (Frenck-Mestre & Pynte, 1997; Juffs & Harrington, 1996). In this section, we
will take a look at second language syntactic processing for ambiguous structures
that are apt to be influenced by the reader’s native language. This question can be
addressed most easily through the study of structures which exist in both the native
and second language but which are processed differently in the two. Just such a case
is provided by the ambiguity illustrated in 5, below.
5a.
5b.
5c.
5d.
5e.
Arnold watched the wife of the doctor who was leaving the medical centre.
La police cherche le fils du médecin qui est soupçonné de trafic de drogues.
Alguien disparó contra la criada del actor que estaba en el balcon.
Patricia conosceva il ragazzo de direttore che era svenuto alla festa.
De gangsters schoten op de zoon van die actrice die op het balkon zat.
Frenck-Mestre
The sentences presented in 5a through 5e all present the same structural
ambiguity
. That is, the subject of the relative clause can be either the head of the complex
noun phrase (i.e. in 5a, "the wife of the doctor") or the second noun phrase alone
(i.e. in 5a, "the doctor"). To date, and despite numerous on-line studies performed in
a wide range of languages, from English to Japanese (for reviews, see Cuetos,
Mitchell & Corely, 1996; Fernandez, 2000; Fodor, 1998; Frazier & Clifton, 1996;
Frenck-Mestre & Pynte, 2000a; 2000b; Gibson, Pearlmutter, Canseco-Gonzalez &
Hicock, 1996), no single model can account for the resolution of this ambiguity.
Factors known to influence the resolving of this ambiguity are, among others, the
respective "weight" of the complex NP and the subsequent relative clause
(Fernandez, 2000; Fodor, 1998; Pynte & Colonna, 2000), the frequency of the two
noun phrases (Frenck-Mestre & Pynte, 2000a; Pynte & Colonna, 2000), the
preposition in the complex NP (Cuetos et al., 1996; De Vincenzi & Job, 1993;
Frenck-Mestre & Pynte, 2000b; Gilboy, Sopena, Clifton & Frazier, 1995),
segmentation of the sentence (Gilboy & Sopena, 1996, but see Carreiras & Clifton,
1999), whether one considers "on-line" or "off-line" measures of noun preference
(De Vincenzi & Job, 1995; Mitchell, Brysbaert & Gondelaers, 1997), and indeed the
language that the ambiguity is presented in (Cuetos et al., 1996; Fernandez, 2000;
Frenck-Mestre, 1997; Frenck-Mestre & Pynte, 2000b; Gibson et al., 1996). It is
this latter factor that is of particular interest for the present discussion.
For purposes of the present paper, we will consider the "simplest" case, that is
where the preposition in the complex NP is "of" or its equivalent in French (and
Spanish, reviewed later) and the printed frequency of the two nouns is either roughly
equivalent or neutralized. Consider the examples provided in 6.
6a. Charles photographie la fille de Marc qui est plus dédaigneu(se/x) que
jamais.
Charles photographed the daughter of Mark who is more disdaining
(fem/masc) than ever.
6b. Aline téléphone aux filles de la gardienne qui reviennent/revient de Paris.
Aline calls the daughters of the nanny who are/is returning from Paris.
6c. Aline téléphone à la gardienne des filles qui revient/reviennent de Paris.
Aline calls the nanny of the girls who is/are returning from Paris.
In 6a, the structure is disambiguated by gender agreement between one of the
nouns in the complex NP and the adjective in the RC. In 6b and 6c, the same
structure is disambiguated by number agreement between NP1 or NP2 and the verb
of the RC. We have shown that in French, the prefered interpretation of these
sentences is to attach the RC to the head of the complex NP, as revealed by the firstpass reading times of native French speakers (Frenck-Mestre, 1997; Frenck-Mestre
& Pynte, 2000b, see also Zagar, Pynte & Rativeau, 1997, but also the debate
between Baccino, De Vincenzi & Job, 2000 and Frenck-Mestre & Pynte, 2000b).
This preference to attach "high" can be accounted for in terms of the "construal"
hypothesis, forwarded by Frazier and Clifton (1996), or the "predicate proximity vs
recency" model proposed by Gibson et al. (1996), or alternatively in the framework
of the "linguistic tuning" hypothesis, advocated by Mitchell and collegues (Cuetos et
al., 1996; Mitchell, 1994; Mitchell, Brysbaert, Grondelaers & Swanepoel, 2000).
These models also offer accounts for why the same structure, presented in English,
is apparently resolved differently by native English speakers. That is, English native
speakers have been shown to either prefer "low" attachment or not to display a
definite preference for either host (Carreiras & Clifton, 1993; 1999; Cuetos &
Mitchell, 1988; Fernandez, 2000). Of interest here, however, is which, if any of
On-line sentence processing in the second language
these models can account for experienced-based differences. That is, for changes in
parsing preferences in the second language of bilingual readers as these readers
become more proficient in their L2.
Indeed, given the reported differences across French and English as concerns the
resolution of this ambiguity, it is of interest to examine how English-French
bilinguals resolve it when reading in French. Furthermore, it is important to
determine whether these bilinguals' parsing preferences will be modified by their
experience with their second language. Indeed, in a beginning stage, it would not be
surprizing to find that English-dominant bilinguals are influenced by their native
language when processing L2 sentences
. What difference, however, will experience with the L2 make? The answer to this
question may lie in the type of structure that is processed, as we will outline shortly.
Consider first, however, the results from a series of recent studies completed in our
centre, with beginning and more advanced English-French bilinguals.
In a first study (reported in Frenck-Mestre, 1997), we asked relatively nonproficient English-French "late" bilinguals (n=16) to participate in a reading
experiment wherein we recorded their eye-movements during the reading of
individually presented French sentences (sentences were presented on a single line).
These bilinguals had a mean of three years of formal learning of French in a
classroom setting outside of France, and nine months of immersion in French, in
France. They rated their abilities as concerns written and oral expression and
comprehension at a level of roughly 5 on a ten point scale of proficiency. The
sentences of interest (n=20) were of the structure illustrated in 6b and c, where the
verb of the subordinate clause agreed in number with either NP1 or NP2. These
sentences were interspersed with three times as many filler sentences of varying
syntactic structures (some ambiguous), such that subjects were not prone to engage
in strategic processing. Subjects were not informed of the ambiguous nature of the
experimental sentences. Native French speakers were included in the study as a
control group. The results, are presented in Figure 1. As can be seen, the two groups
showed a quite different pattern of results as concerns the immediate resolution of
this ambiguity. Native French speakers showed a definite preference for "high"
attachment of the RC. That is, first pass (as well as total) reading times were
significantly faster at the disambiguating subordinate verb when it agreed with the
head of the complex NP than when it agreed with NP2. The opposite was observed
in the group of English-dominant bilinguals. These readers, who had been living in
France for a mean of nine months and had a mean of three years of formal classroom
learning in French, showed a trend toward N2 attachment. That is, their first pass
(and total) reading times tended to be faster when the subordinate verb agreed in
number with the second than with the first noun of the complex NP
. As we have argued elsewhere (cf. Frenck-Mestre, 1997), the pattern obtained in
the English-dominant group is to be attributed most likely to the influence of their
native language on second language processing, rather than to a general strategy of
attaching new elements to the most recently processed constituant.
Frenck-Mestre
Figure 1. First pass gaze durations as a function of type of reader and sentence
region for temporarily ambiguous relative clause constructions.
More recently, (Frenck-Mestre, 1998; submitted), we examined the performance
of proficient English-French "late" bilinguals for the same ambiguous structure in a
new eye-movement experiment. The same set of sentences and same presentation
were employed as in the previous experiment. The group had roughly three years of
formal learning of French outside of the country, at least two years of study in a
French university in a mainstream curriculum, and had been living in France for a
mean of five years. They rated themselves at a level of 7 or better on a ten point
scale of proficiency in the second language for written and oral comprehension and
production skills. The results of these "proficient" bilinguals are compared to the
performance of the same group of native French speakers examined in the first
study, in Figure 2.
-----------------------------------------------
On-line sentence processing in the second language
Figure 2. First pass gaze durations as a function of type of reader and sentence
region for temporarily ambiguous relative clause constructions.
As is visible in the graph, the proficient bilinguals' syntactic processing of the
NP1-Prep-NP2-RC ambiguity in French was highly similar to that of native French
speakers. In fact, the statistical comparisons of these two groups did not reveal any
significant interactions with type of reader. The proficient bilinguals were not slower
in their first pass reading times than native French speakers (p>.10), nor did they
demonstrate a different pattern of ambiguity resolution than native speakers.
A comparison was made of the three groups – native French, beginning and
advanced English-French bilinguals – as concerns the first pass gaze duration at the
disambiguating subordinate verb. This comparison is presented in Figure 3. As can
be readily seen, the native and advanced groups showed a highly similar pattern of
ambiguity resolution for the NP1-Prep-NP2-RC structure, which differed from the
group of beginning (English-dominant) bilinguals. This was born out in statistical
analyses, showing an interaction between type of reader and attachment preference
(F1(1,36) = 4.89, p<.01). Post-hoc comparisons revealed that the beginning
bilinguals differed from the advanced and native groups as concerns the time spent
reading the subordinate verb in the case of "high" i.e. N1 attachment. The beginning
bilinguals were significantly slower to process the subordinate verb than advanced
bilinguals and native speakers when the verb agreed in number with the first noun
phrase. Otherwise stated, these English-dominant bilinguals demonstrated specific
difficulty with "high" attachment.
Frenck-Mestre
Figure 3. Initial reading times at the disambiguating verb of the relative clause as a
function of type of reader.
Our data show a clear evolution of processing strategies, from performance
closely tied to the native language of late bilinguals to that closely resembling the
performance of native speakers of the language. It is important to underline that the
difference between the two groups of bilinguals was experience with the second
language. The advanced group had lived for several years in the country of their
second language, in environments where they used the language as a daily means of
communication. They had been "exposed" to their second language, and its patterns,
for several years. Prior to examining which model of processing can best account
for these data, we will consider the results of another study of the same syntactic
ambiguity.
Our data can be compared to those reported recently by Fernandez (1998), for
Spanish-English and Japanese-English bilinguals, in off-line studies of the same,
relative clause attachment ambiguity. It is clear from the questionnaire data
provided in this study that the deciding factor behind ambiguity resolution in the
second language was level of expertise in this language. Whether the bilinguals
learned their second language early or late was not predictive of the amount of
forward transfer from the bilinguals' native language. Self-rated proficiency,
however, was. Those Spanish speakers who rated themselves as more proficient in
English than in Spanish processed this ambiguity akin to native English speakers,
whereas those who rated themselves as more proficient in Spanish tended to process
this ambiguity as would monolingual Spanish speakers. In like manner, the Japanese
bilinguals, despite having learned English fairly, rated themselves in general as
"Japanese dominant" and showed clear effects of transfer from their native language.
In our study, the number of years of experience with and/or immersion in the second
language (French) was the deciding factor : less proficient bilinguals showed clear
effects of forward transfer whereas the proficient group was well on its way to
processing the ambiguity in question as would a monolingual.
These data clearly show that any processing model of L2 syntactic analysis must
take into account the evolution of processing with experience. The most readily
adapted framework for this would seem to be the "linguistic tuning" hypothesis
(Cuetos et al., 1996; Mitchell, 1994). The model incorporates both cross-linguistic
differences in ambiguity resolution and changes over time in processing as a result
On-line sentence processing in the second language
of exposure, due to the very nature of the model. Indeed, it is assumed that the
particular anaylsis adopted by the adult reader will be the product of his/her
exposure to a given language and its statistical properties as concerns the frequency
of structures. Changes in syntactic processing are predicted as a result of exposure,
as clearly stated in Cuetos et al. (1996, p. 175) "The model predicts that parsing
preferences will change if, during some period prior to testing, the reader or listener
has been exposed to an unusual preponderance of one ambiguity resolution rather
than another." Indeed, our data (as well as Fernandez') show that English-French
bilinguals will adopt a "French" strategy of "high attachment" for the RC-attachment
ambiguity after having been exposed to French for numerous years
.
The linguistic tuning hypothesis is not alone, however, in being able to account
for cross-linguistic variation as concerns the ambiguity under question, nor, as we
will outline, for the evolution of processing we observed in our bilingual subjects as
concerns this structure. The construal hypothesis, forwarded by Frazier and Clifton
(1996) stipulates that 'non-primary' relationships such as relative clauses are
'construed' rather than being governed by strictly syntactic principles. Attachment
would be the result, in this case, of "late" discourse level processes. This leaves open
the possibilty for cross-linguistic variation. Moreover, given the apparently strong
preference for "high" attachment of the RC following a genetive relationship ("the
daughters of the nanny") in French (and Spanish), it is perhaps not surprizing that
native English speakers would adhere to this processing strategy, all the more so as
the preference in English is apparently not as clear-cut.
Conclusions
In summary, the present review of a handful of on-line studies of second
language syntactic processing has shown, first, that second language syntactic
analysis need not be considered as a separate case from native-language parsing.
Proficient bilinguals show immediate sensitivity to structural ambiguities in like
manner to native readers, quite independently of reading speed. These proficient
bilinguals are also able to take advantage of fine-levels of analysis to guide (or filter)
the parsing of structurally ambiguous sentences in the second language. Second, in
the instance that the bilingual's two languages differ as concerns the processing of a
given syntactic structure, changes in parsing strategies will occur over time as the
bilingual gains experience with the second language. A decrease in forward transfer
is observed for more proficient bilinguals along with increased "native-like"
processing.
References
Altmann, G. & Steedman, M. (1988). Interaction with context during human
sentence processing. Cognition, 30, 191-238.
Baccino, T., DeVincenzi, P. & Job, R. (2000). Cross-linguistic studies of the late
closure strategy: French and Italian. In M. De Vincenzi & R. Job (Eds). Crosslinguistic perspectives on language processing. Kluwer Academic Press.
Boland, J.E. & Boehm-Jernigan, H. (1998). Lexical constraints and prepositional
phrase attachment. Journal of Memory and Language, 39, 684-719.
Carreiras, M. & Clifton, C. (1993). Relative clause interpretation preferences in
Spanish and English. Language and Speech, 36, 353-372.
Frenck-Mestre
Carreiras, M. & Clifton, C. (1999). Another word on parsing relative clauses: Eyetracking evidence from Spanish and English. Memory and Cognition, 27, 5, 826833.
Clifton, C., Speer, S. & Abney, (1991). Parsing arguments: Phrase structure and
argument structure as determinants of initial parsing decisions. Journal of Memory
and Language, 30, 251-271.
Cuetos, F. & Mitchell, D.C. (1988). Cross-linguistic differences in parsing:
Restrictions on the late-closure strategy in Spanish. Cognition, 30, 73-105.
Cuetos, F., Mitchell, D.C. & Corley, M. (1996). Parsing in different languages. In
M. Carreiras, J. Garcia-Albea and N. Sebastien-Galles (Eds.) Language Processing
in Spanish Mahwah, NJ: LEA
DeVincenzi, P. & Job, R. (1993). Some observations on the universality of the late
closure strategy. Journal of Psycholinguistic Research, 22, 189-206.
DeVincenzi, P. & Job, R. (1995). An investigation of late closure: The role of
syntax, thematic structure and pragmatics in initial and final interpretations. Journal
of Experimental Psychology: Learning, Memory and Cognition. 21, 1303-1321.
Durgunoglu, A.Y. & Hancin, B.J. (1992). An overview of cross-language transfer
in bilingual reading. In R.J. Harris (Ed.), Cognitive Processing in Bilinguals.
Amsterdam: Elsevier, North Holland
Fernandez, F. (1998). Language Dependency in Parsing : Evidence from
monolingual and bilingual processing. Psychologica Belgica, 38, 197-230.
Fernandez, F. (2000). Bilingual sentence processing : Relative clause attachment in
English and Spanish. Unpublished Doctoral Dissertation. CUNY.
Ferrand, L. & Grainger, J. (1993). The time course of orthographic and
phonological code activation in the early phases of visual word recognition. Bulletin
of the Psychonomic Society, 31, 119-122.
Ferreira, F. & Henderson, J.M. (1990). Use of verb information in syntactic
processing: Evidence from eye movements and word-by-word self-paced reading.
Journal of Experimental Psychology: Learning, Memory and Cognition, 16, 555568.
Flynn, S. (1984). A universal in L2 acquisition based on a PBD typology. In F.
Eckman, L. Bell & D. Nelson (Eds) Universals of Second Language Acquistion.
Rowley, Mass: Newbury House
Flynn, S. & Espinal, I. (1985). Head initial/head final parameter in adult Chinese L2
acquisition of English. Second Language Research, 1, 93-117.
Fodor, J. (1998). Learning to parse? Journal of Psycholinguistic Research, 27, 285319.
Frazier, L. (1987). Sentence processing: A tutorial review. In M. Coltheart (Ed).
Attention and performance XII. Hillsdale, NJ: LEA
Frazier, L. (1990). Parsing modifiers: Special purpose routines in HSPM? In D.A.
Balota, G.B. Flores d'Arcais & K. Rayner (Eds.), Comprehension Processes in
Reading. Hillsdale, NJ: LEA
Frazier, L. & Clifton, C. (1996). Construal. Cambridge, MA.: MIT Press.
Frenck-Mestre,C. (1993). Use of orthographic redundancies and word identification
speed in bilinguals. Journal of Psycholinguistic Research. 22, 397-410.
Frenck-Mestre, C. (1994). Native language versus second language reading: Where
does the difference lie? Paper presented at the VII European Society for Cognitive
Psychology. Lisbon, Portugal. 10-14 sept
Frenck-Mestre, C. (1997). Examining second language reading: An on-line look. in
A. Sorace, C. Heycock and R. Shillcock (Eds.) Language Acquisition, Knowledge
Representation and Processing: GALA 1997. HCRC, Edinburgh.
Frenck-Mestre, C. (1998). Overcoming parameters : An on-line look at bilingual
sentence processing. Paper presented at the 39th annual meeting of the Psychonomic
Society. Dallas, 18 – 22 ,nov.
Frenck-Mestre, C. & Pynte, J. (1995). Lexical influences on parsing strategies:
On-line sentence processing in the second language
Evidence from eye movements. In J.M. Findlay, R.W. Kentridge & R. Walker
(Eds.) Eye movement research: Mechanisms, Processes and Applications.
Amsterdam: Elsevier, North Holland.
Frenck-Mestre, C. & Pynte, J. (1997). Syntactic ambiguity resolution while reading
in second and native languages. The Quarterly Journal of Experimental Psychology,
50A, 119-148.
Frenck-Mestre, C. & Pynte, J. (2000a). Romancing syntactic ambiguity: Why the
French and the Italians don't see eye to eye. In A. Kennedy, R. Radach, D. Heller
and J. Pynte (Eds.) Perceptual Processes in Reading. DeBoeck
Frenck-Mestre, C. & Pynte, J. (2000b). Resolving syntactic ambiguities: Crosslinguistic differences? In M. deVincenzi and V. Lombardo (Eds) Cross-linguistic
perspectives on language processing. Kluwer Academic Press. 119-148.
Frost, R., Katz, L. & Bentin, S. (1987). Strategies for visual word recognition and
orthographical depth: A multilingual comparison. Journal of Experimental
Psychology: Human Perception and Performance. 13, 104-115.
Gibson, E., Pearlmutter, N., Canseco-Gonzales, E. & Hickok, G. (1996). Crosslinguistic attachment preferences: Evidence from English and Spanish. Cognition,
59, 23-59.
Gilboy, E., & Sopena, J.M. (1996) Segmentation effects in processing complex noun
pronouns with relative clauses. In M. Carreiras, J. Garcia-Albea and N. SebastienGalles (Eds.) Language Processing in Spanish Mahwah, NJ: LEA
Gilboy, E., Sopena, J.M., Clifton, C. & Frazier, L. (1995). Argument structure and
association preferences in Spanish and English complex NPs. Cognition, 54, 131167.
Hoover, M.L. & Dwivedi, V.D. (1998). Syntactic processing by skilled bilinguals.
Language Learning, 48, 1-29.
Juffs, A. & Harrington, M. (1996). Garden path sentences and error data in second
language sentence processing. Language Learning, 46, 283-326.
Koda, K. (1993). Transferred L1 strategies and L2 syntactic structure in L2
sentence compre-hension. The Modern Language Journal, 77, 490-499.
MacWhinney, B. (1992). Transfer and competition in second language learning. in
R.J. Harris (Ed.), Cognitive Processing in Bilinguals. Amsterdam: Elsevier, North
Holland.
Meisel, J., Clahsen, H. & Pienemann, M. (1981). On determining developmental
stages in natural second language acquisition. Studies in Second Language
Acquisition, 3, 109-135.
Mitchell, D.C. (1994). Sentence parsing. In M.A. Gernsbacher (Ed.), Handbook of
Psycholinguisitics. New York: Academic Press.
Mitchell, D.C. (1989). Verb-guidance and other lexical effects in parsing. Language
and -Cognitive Processes, 4, 123-154.
Mitchell, D.C. & Holmes, V.M. (1995). The role of specific information about the
verb in parsing sentences with local structural ambiguities. Journal of Memory and
Language, 24, 542-549.
Muysken, P. (1984). The Spanish that Quechua speakers learn. In R. Anderson (Ed.)
Second Languages: A Cross-Linguistic Perspective. Rowley, Mass: Newbury House
Odlin, T. (1989). Language Transfer. Cross linguistic influence in language
learning. Cambridge: Cambridge University Press.
Paap, K. McDonald, J.E., Schvaneveldt, R.W. & Noal, R.W. (1987). Frequency and
pronounceability in visually presented naming and lexical decision tasks. In M.
Coltheart (Ed.) Attention and Performance XII (pp. 221-244), Hillsdale, NJ:
Erlbaum.
Pynte, J. & Colonna, S. (2000). Decoupling syntactic parsing from visual
inspection. The case of relative clause attachment in French. In A. Kennedy, R.
Radach, D. Heller and J. Pynte (Eds.) Perceptual Processes in Reading. DeBoeck
Rayner, K. Carlson, M. & Frazier, L. (1983). The interaction of syntax and
Frenck-Mestre
semantics during sentence processing: Eye movements in the analysis of
semantically biased sentences. Journal of Verbal Learning and Verbal Behavior, 22,
358-374.
Rayner, K., Sereno, S.C., Morris, R.K., Schmauder, A.R. & Clifton, C. (1989). Eye
movements and on-line language comprehension processes. Language and Cognitive
Processes, 4, 21-49.
Segalowitz, N. (1986) Skilled reading in the second language. In J. Vaid (Ed.)
Language processing in bilinguals: Psycholinguistic and neuropsychological
perspectives. Hillsdale, NJ: Erlbaum.
Seglaowitz, N. & Hébert,M. (1990). Phonological recoding in the first and second
language reading of skilled bilinguals. Language Learning, 40, 503-538.
Sweet, H. (1899/1972). The practical study of languages. Language and language
learning series. London: Oxford University Press.
Taraban, R. & McClelland, J.L (1988). Constituent attachment and thematic role
assignment in sentence processing: Influences of content-based expectations.
Journal of Memory and Language, 27, 1-36.
Trueswell, J.C., Tanenhaus, M.K. & Kello, C. (1993). Verb-specific constraints in
sentence processing: Separating the effects of lexical preference from garden paths.
Journal of Experimental Psychology: Learning, Memory and Cognition, 19, 528553.
Weber-Fox, C.M. & Neville, H. (1996). Maturational constraints on functional
specializations for language processing: ERP and behavioral evidence in bilingual
speakers. Journal of Cognitive Neuroscience, 8, 231-256.
Zagar, D., Pynte, J. &é Rativeau, S. (1997). Evidence for early closure attachment
on first-pass reading times in French. Quaterly Journal of Experimental Psychology,
50A, 421-438.
Zobl, H. (1986). Word order typology, lexical government and the prediction of
multiple, graded effects in L2 word order. Language Learning, 36, 159-183.
On-line sentence processing in the second language
Notes de bas de page
Note, there were differences betweeen the fast and slow L2 readers at the final
segment of sentences, whereby the latter group was significantly slower for
strucurally ambiguous sentences. Hoover and Dwivedi (1998) attributed this to the
slower readers' lesser ability to process these structures.
Note, there were also effects of inappropriate forward transefer, from the native to
second language as concerns lexical constraints, discussed in Frenck-Mestre &
Pynte (1997).
Don Mitchell provided an impressive list of the studies run to date, across more
than a dozen languages, including both major European languages, Japanese,
Africaans and more, at a recent AMLaP meeting (1998).
Since the end of the nineteenth century, linguists have provided us with examples of
"transfer" whenever two languages come into contact (Jespersen, 1912; Palmer,
1917; Sweet, 1899/1972). The sentences produced by non-native speakers often bear
the mark of their first language (Durgunoglu & Hancin, 1992; Koda, 1993; Odlin,
1989; MacWhinney). For example, word order in second language sentences is
frequently affected by native language constraints (Meisel, Clahsen & Pienemann,
1981). Nonetheless, the results of various studies have underlined cross-linguistic
commonalities in second language acquisition as concerns syntactic development,
which support the hypothesis of linguistic universals in second language
development (Muysken, 1984; Zobl, 1986), rather than a "cognitive-based"
approach to second language learning. At the intersection of these two theoretical
viewpoints, Flynn and collaborators (Flynn, 1984; Flynn & Espinal, 1985) highlight
the important role played by the native language of adult learners in the acquisition
of second language syntax while underlining the universal nature of acquisition.
In both the native French and the beginning bilingual group, the effect of
attachment was independent of whether the subordinate verb was singular or plural.
In a recent ERP study of early and late bilinguals (Weber-Fox & Neville, 1996), it
was found that "late" bilinguals do not show the same sensitivity as early bilinguals
or monolinguals to phrase structure violations. It would be of interest to see
whether, eye-movement recordings would replicate these findings, or, rather, if the
reported effect is linked to the inability to "repair" for this type of sentence.
It is worth noting that, though attractive, the linguistic tuning hypothesis has
experienced what would appear to be a serious set-back in recent studies involving
Dutch (Mitchell et al., 1997; Mitchell & Brysbaert, 1998). That is, when on-line RC
attachment is examined for materials modelled closely on corpora-based sentences,
the on-line preferences are quite the opposite that which would be predicted by the
statistical frequency of structures in written Dutch.